Модель Корнельского университета прогнозирует победу демократов в Палате представителей США с вероятностью 80%

Democrats have an 8-in-10 chance of winning control of the U.S. House this fall, according to a Cornell University forecasting model that projects a roughly 226-209 seat split. The model covers all 435 districts and uses information collected at least 100 days before the Nov. 3 election. Its simulations produced outcomes ranging from 206 to 258 Democratic seats, underscoring the uncertainty around the central estimate. Peter Enns, a professor of government and public policy, presented the forecast Sept. 3 at the American Political Science Association's annual meeting in Boston with doctoral students Leigh Farah, Thomas Gareau-Paquette and Claudia Miner. A Senate forecast is still being developed. Model combines district and national signals The system considers how each district voted previously, expert assessments, whether a race is competitive or uncontested, campaign donations and incumbent status. It also weighs state and national polling to capture the broader political mood. Enns said voter-intention data collected 100 days before Election Day showed more support for Democrats than two years earlier. The president's party also historically tends to lose seats in midterm elections. Unlike models that focus mainly on state and national indicators, the Cornell approach forecasts every House district and reflects recent redistricting. Historical tests used only information that would have been available 100 days before each election. Historical tests show high accuracy The researchers ran thousands of simulations to create a 95% confidence interval. Across more than 6,500 races since 1996, the model achieved 96% overall accuracy and correctly forecast two-thirds of contests experts rated as toss-ups, Cornell said. The model has correctly identified the winning party in the previous 14 congressional elections. Enns also forecast every state's presidential winner and the Electoral College total in 2024, and missed only Georgia in 2020. Enns cautioned that the forecast is not deterministic. He said a Republican House majority remains possible but would require an unusually favorable combination of outcomes or a break from the patterns captured by the model. Forecasts can explain why races turn Researchers use advance forecasts not only to predict control of Congress but to test which factors help explain election results. Strong performance by a variable such as campaign donations, for example, could signal candidate popularity or the effect of increased advertising, though more study would be needed to distinguish the cause. Comparing the forecast with actual results can also show whether races shifted in the closing weeks or were largely settled earlier. District-level differences could help researchers evaluate the effects of the economy, presidential approval, mail-voting rules, candidate ideology and redistricting. Enns said results close to the model's estimates would suggest the election followed the patterns seen since at least 1996. Larger departures could point researchers toward events or conditions the historical data did not anticipate.